Glomerular Filtration and Renal Blood Flow

At rest, an adult human’s two kidneys, well under 1% of body mass, receive about a quarter of the blood the heart pumps, and they extract little of its oxygen.

That flow supplies a filter that passes about 125 mL of plasma a minute in an adult at rest. The filter lets water and small solutes through and holds back plasma protein, and how much it passes depends on net filtration pressure, the ultrafiltration coefficient and the resistance of the afferent and efferent arterioles on either side of the capillary. Autoregulation, acting mainly on the afferent arteriole, holds the rate steady when arterial pressure changes. The kidney traced the nephron and introduced clearance, C = U x V / P. This lesson puts numbers on its pressures, flows and controls.

A capillary that stays at high pressure

Unlike most systemic capillaries (capillary exchange and lymph), the glomerular capillary sits between two resistances, the afferent and efferent arterioles, and its hydrostatic pressure barely falls along its length.

Four Starling forces set net filtration pressure. Glomerular capillary hydrostatic pressure (P_GC) and the oncotic pressure in Bowman’s space (pi_BS, normally near zero) favour filtration. Bowman’s-space hydrostatic pressure (P_BS) and capillary plasma oncotic pressure (pi_GC) oppose it.

SNGFR = Kf x (P_GC - P_BS - pi_GC + pi_BS), averaged along the capillary

SNGFR is single-nephron GFR. Kf, the ultrafiltration coefficient, is filtration per unit net pressure: water permeability times filtering area. The equation treats the wall as almost impermeable to protein. As water leaves and protein stays, pi_GC rises while P_GC stays nearly flat, so net pressure shrinks along the capillary.

The direct numbers come from rats. Brenner, Troy and Daugharty (1971) used Munich-Wistar rats, which have some glomeruli at the kidney surface, and measured P_GC with a servo-nulling micropipette: about 60 cmH2O (about 44 mmHg), roughly half of mean arterial pressure, with net filtration pressure averaging about 14 cmH2O (about 10 mmHg). In hydropenic rats (mildly water-deprived, so not volume-expanded), oncotic pressure typically rose to equal the hydrostatic difference before the efferent end: filtration pressure equilibrium. Past that point nothing filters, so more plasma flow lengthens the filtering stretch and SNGFR depends strongly on flow. Human P_GC is inferred rather than measured this way, and whether human glomeruli reach equilibrium is unknown.

The filter has three layers. Fenestrated endothelium is lined by a negatively charged glycocalyx. The basement membrane lies between the endothelium and the podocytes. Size selectivity comes mostly from the cell layers. Podocytes wrap the capillary in foot processes bridged by the slit diaphragm. Human genetics implicates it: NPHS1 mutations (nephrin) cause congenital nephrotic syndrome of the Finnish type, and NPHS2 mutations (podocin) segregate with steroid-resistant nephrotic syndrome, both with heavy protein loss in urine. Basement-membrane charge was long taught as the albumin barrier, but mice with reduced membrane charge did not develop overt proteinuria. Whether charge elsewhere in the barrier, such as the glycocalyx, helps hold back albumin is still debated.

Clearance measures the whole kidney

Clearance gives the sum over both kidneys. If a substance is filtered and neither reabsorbed nor secreted, U x V = GFR x P, and its clearance is the filtration rate. In 1926 Poul Brandt Rehberg, in August Krogh’s Copenhagen laboratory, tested this on himself with creatinine, and a 2025 review summarises his result as about 180 L of filtrate a day; with 1 to 2 L of urine, about 99% of filtered water must be reabsorbed. That made filtration with massive reabsorption quantitatively workable in one human. His result did not show that creatinine is only filtered, as the worked problem below and Going deeper explain. Homer Smith’s work led to inulin (Shannon and Smith, 1935) and to para-aminohippurate, PAH, for plasma flow (Smith and colleagues, 1945); see osmoregulation.

PAH is filtered and secreted, and one pass removes roughly 90% of it from arterial blood, so its clearance is effective renal plasma flow (ERPF). Dividing by the extraction gives renal plasma flow (RPF); dividing by the plasma fraction of blood gives renal blood flow (RBF). The filtration fraction (FF) is the fraction of glomerular plasma flow that is filtered.

RPF = C_PAH / E_PAH; RBF = RPF / (1 - Hct); FF = GFR / RPF

Worked problem: filtration, flow and a drifting marker

Made-up round numbers, adult human at rest, steady state, both kidneys. Assume inulin is freely filtered and neither reabsorbed nor secreted, PAH extraction is 0.90 (assumed), hematocrit 0.45 and cardiac output 5.0 L per min.

1. GFR. U_inulin = 25 mg per mL, V = 1.0 mL per min, P_inulin = 0.20 mg per mL. C = 25 x 1.0 / 0.20 = 125 mL per min, or 125 x 1,440 = 180 L per day, the size of Rehberg’s estimate.
2. Flows. U_PAH = 12.5 mg per mL, P_PAH = 0.020 mg per mL. ERPF = 12.5 x 1.0 / 0.020 = 625 mL per min. RPF = 625 / 0.90 = 694 mL per min. RBF = 694 / 0.55 = about 1,260 mL per min, about 25% of cardiac output.
3. Filtration fraction. 125 / 694 = 0.18, or 125 / 625 = 0.20 using ERPF.
4. Along one capillary (illustrative, human-like). P_GC = 55 mmHg throughout, P_BS = 15, afferent pi_GC = 25, pi_BS = 0 (all mmHg). Afferent end: 55 - 15 - 25 = +15 mmHg. Because 18% of the plasma filters out, the remaining 82% holds the same protein, so protein is concentrated 1 / (1 - 0.18) = 1.22-fold by the efferent end; if oncotic pressure were proportional to concentration, pi_GC = 30.5 mmHg and net pressure = +9.5 mmHg. Real oncotic pressure rises faster than linearly, so this underestimates efferent pi_GC.
5. Creatinine. Take production as a constant 1.0 mg per min (illustrative) and creatinine as filtered only. At steady state excretion equals production, so P_cr = 1.0 / 125 = 0.008 mg per mL = 0.8 mg per dL. If GFR halves, P_cr rises over days to 1.6 mg per dL, when excretion again matches production.

But creatinine is also secreted. Proximal tubule cells also secrete creatinine through transporters, so creatinine clearance exceeds GFR, and a drug that blocks those transporters, such as cobicistat, raises plasma creatinine without changing GFR. Creatinine-based estimates also depend on muscle mass, and scaling GFR to 1.73 m2 of body surface is a convention.

Two resistors in series: afferent and efferent arterioles

Think of a hose with a tap upstream (afferent) and a nozzle downstream (efferent). RBF = (arterial pressure - renal venous pressure) / (afferent + efferent resistance), and P_GC is arterial pressure minus the drop across the afferent arteriole (resistance, compliance and local flow).

• Afferent constriction: RPF and P_GC fall, so GFR falls; in the simple model FF changes little. Dilation does the reverse.
• Efferent constriction: RPF falls, P_GC and FF rise, and GFR is biphasic. Mild constriction raises GFR slightly. Severe constriction lowers RPF and GFR, because the rise in oncotic pressure outruns the rise in hydrostatic pressure.
• Efferent dilation: RPF rises; P_GC, GFR and FF fall.

Afferent resistance ×1.00 (baseline ×1.00) Efferent resistance ×1.00 (baseline ×1.00) Arterial pressure 100 mmHg (baseline 100) Autoregulation on (a rule assumed for this sketch)

GFR

Renal plasma flow, RPF

Renal blood flow, RBF

Filtration fraction, FF

Glomerular capillary pressure, P_GC

Net pressure, afferent end then efferent end

In this sketch: steady state, adult human at rest, both kidneys. Blood flow Q = (arterial pressure − renal venous pressure) / (afferent resistance + efferent resistance), ignoring the volume filtered in the series resistance, and P_GC = arterial pressure − Q × afferent resistance. The capillary is cut into 100 segments; each filters Kf / 100 × (P_GC − P_BS − pi) and stops filtering when its net pressure is not positive. Pi rises in proportion to protein concentration (a linear rule: the lesson notes real oncotic pressure rises faster than linearly, so this underestimates pi downstream), and there are no feedback dynamics. From the lesson’s worked problem (made-up round numbers): P_BS 15 mmHg, afferent pi 25 mmHg, hematocrit 0.45, P_GC 55 mmHg, GFR 125 mL per min, RPF 694 mL per min, FF 0.18. Made up for the sketch: renal venous pressure 5 mmHg, an arterial pressure of 100 mmHg at baseline, the split of baseline resistance between the two arterioles (the one that gives P_GC 55 mmHg), Kf (chosen so baseline GFR is 125 mL per min), the slider ranges, the autoregulation rule (afferent resistance rescaled so that P_GC stays at 55 mmHg over 80 to 180 mmHg when the efferent resistance is at baseline, a stand-in for the lesson’s “about 80 to 180 mmHg”) and a 1 mmHg tolerance for calling net pressure zero. In this sketch the size of the FF change with afferent resistance and of the GFR rise and fall with efferent resistance come from these made-up numbers and do not match the lesson’s “changes little” and “slightly”. All numbers are schematic, and arteriole thickness is drawn thinner for higher resistance, not to scale.

Rule to keep: the two arterioles are resistors in series, so P_GC is arterial pressure minus the drop across the afferent arteriole; as water leaves the capillary and protein stays, oncotic pressure rises and net filtration pressure shrinks along the capillary.

When perfusion pressure falls, angiotensin II is described as acting preferentially on efferent AT1 receptors, supporting GFR better than plasma flow. Blocking its formation (ACE inhibitors) or receptor (ARBs) dilates the efferent arteriole and can lower GFR, most when perfusion pressure is low. Prostaglandins dilate the afferent arteriole; NSAIDs, which block their synthesis, can let it constrict and lower GFR, most when angiotensin II and other constrictors are high.

Holding GFR steady: myogenic response and tubuloglomerular feedback

In mammals RBF and GFR stay nearly constant over renal perfusion pressures of about 80 to 180 mmHg: autoregulation. Two mechanisms, both acting mainly on the afferent arteriole, do most of it.

In the myogenic response, stretch of arteriolar smooth muscle depolarizes it, cytosolic calcium rises and it contracts. Bayliss described the reaction in arteries in 1902 (smooth muscle); the stretch sensor is still debated.

Tubuloglomerular feedback runs through the macula densa, a plaque of cells at the end of the thick ascending limb where the tubule touches its own glomerulus. NaCl enters them through NKCC2, the cotransporter loop diuretics block. When NaCl delivery rises, the cells release ATP; adenosine formed from it constricts the afferent arteriole through adenosine A1 receptors (A1 is Gi-coupled; G proteins), and P_GC and SNGFR fall; in some preparations ATP may also act directly on P2X receptors. When delivery falls, the afferent arteriole dilates and granular cells release renin.

Sun and colleagues in Schnermann’s group (2001) used micropuncture in anesthetized wild-type mice and mice lacking A1 receptors. They perfused the loop of Henle at 0 and then 30 nl per min and recorded proximal stop-flow pressure, which follows P_GC. If A1 receptors carry the signal, stop-flow pressure should fall in wild type only. It fell 6.8 mmHg in wild type and 0.4 mmHg in knockouts; baseline GFR, blood pressure and heart rate did not differ. So A1 receptors are required for the feedback response in these mice, and a mediator bypassing A1 signalling is ruled out in this preparation. It does not show where the adenosine comes from or that feedback sets resting GFR, and tests no human; the knockout is global, and development may compensate.

Diabetes gives a second test. High plasma glucose raises filtered glucose, and the proximal tubule reabsorbs more of it with sodium through SGLT2. Less NaCl reaches the macula densa, the afferent arteriole dilates, and P_GC and GFR rise: hyperfiltration. (Whether blunted feedback or other preglomerular dilation dominates is open.) Blocking SGLT2 should reverse the chain. In rats with early streptozotocin diabetes, dapagliflozin raised early distal chloride by about 70% and sharply lowered SNGFR. In Zucker diabetic fatty rats, luseogliflozin reduced afferent dilation without changing efferent diameter, and an A1 antagonist blocked the effect. In adults with type 1 diabetes under clamped glycemia, Cherney and colleagues (2014) measured inulin and PAH clearance before and after eight weeks of empagliflozin; in the high-GFR group GFR fell from about 172 to 139 mL per min per 1.73 m2, ERPF fell and renal vascular resistance rose. That study had no control arm, and clearances cannot say which arteriole changed. SGLT2 at the tubule is in tubular transport and diuretics.

Design problem

Filtration pressure equilibrium hides Kf. You study one surface glomerulus in an anesthetized, hydropenic Munich-Wistar rat, as Brenner’s group did. Pipettes give P_GC and tubule pressure, arterial and efferent arteriolar blood give oncotic pressures, and a timed proximal collection gives SNGFR. Made-up values: SNGFR 30 nl per min, P_GC 45 mmHg, P_BS 10 mmHg, afferent oncotic pressure 19 mmHg, efferent 35 mmHg. What can you say about Kf? Design a way to measure it.

One way to do it

SNGFR = 0.50 nl per s. The hydrostatic difference is 45 - 10 = 35 mmHg, so net pressure is 16 mmHg at the afferent end and 0 at the efferent end. Equilibrium was reached at an unknown point. If net pressure falls fastest where it is highest, as the usual curves show, mean net pressure is at most about (16 + 0) / 2 = 8 mmHg. So Kf is at least 0.50 / 8 = 0.0625 nl per s per mmHg: a minimum, not a value.

Move the capillary out of equilibrium. Raise glomerular plasma flow, for example with an infusion of rat plasma, until efferent oncotic pressure stays below 35 mmHg, then integrate net pressure along the capillary using a model of oncotic pressure against protein concentration. Illustrative result: SNGFR 0.80 nl per s, net pressure 19 mmHg falling to 7 mmHg, integrated mean 10 mmHg, so Kf = 0.80 / 10 = 0.08 nl per s per mmHg, the value Deen and colleagues (1973) reported. Repeat at a second plasma flow: Kf should not change, and SNGFR’s rise with flow should come from mean net pressure, as they found.

Limits: compare the same rats before and after; anesthesia and surgery alter hemodynamics; surface glomeruli may not represent deep nephrons; this is a rat value. Do not infer filtration fraction from the oncotic ratio, because oncotic pressure is not proportional to protein concentration.

Going deeper

Where the filtrate goes: kidney, ureter and bladder

Each kidney has an outer cortex, where the glomeruli lie, and an inner medulla of loops and collecting ducts, which drain into the renal pelvis and the ureter. Peristalsis carries urine down the ureter. Each ureter crosses the bladder wall obliquely, so rising bladder pressure presses it shut and limits backflow.

The bladder wall, the detrusor, is smooth muscle with features of both single-unit and multi-unit types (smooth muscle). The internal urethral sphincter is smooth muscle at the bladder neck, not under voluntary control (a distinct ring in males; in females the bladder neck and proximal urethra act as a functional sphincter); the external urethral sphincter is voluntary skeletal muscle.

During storage, sympathetic fibres in the hypogastric nerve relax the detrusor through beta-3 adrenoceptors and contract the bladder neck through alpha-1 adrenoceptors, while the pudendal nerve keeps the external sphincter contracted. Voiding is a brainstem reflex, worked out mainly in cats and supported in humans by imaging: bladder stretch afferents reach the periaqueductal gray, which relays to the pontine micturition centre (Barrington’s nucleus). Its output excites sacral preganglionic neurons, conventionally called parasympathetic (a label now debated), whose pelvic nerve fibres excite neurons in pelvic and bladder-wall ganglia; these postganglionic neurons release acetylcholine, which contracts the detrusor through muscarinic M3 receptors, while sympathetic and pudendal outflow are inhibited and both sphincters relax. Cortical control of the reflex matures in childhood.

How clearance was worked out

The logic of clearance was worked out in a live dispute: does urine begin as a filtrate that the tubules mostly reabsorb (the line from Carl Ludwig), or do kidney cells secrete it (the line of William Bowman and Rudolf Heidenhain)? Arthur Cushny’s 1917 book argued for filtration and reabsorption, and Heidenhain’s urea-based argument that about 70 L of filtrate a day would have to be reabsorbed already seemed a large objection.

Rehberg took a load of creatinine by mouth and measured it in plasma and timed urine hourly. Urea is partly reabsorbed, so its clearance runs low; creatinine is concentrated in urine far more, so it should come closer to the filtration rate. His clearance varied between periods, from about 50 to 200 mL per min. His result did not show that creatinine is only filtered; Rehberg could not exclude tubular handling, and a 2023 history argues that the high plasma creatinine from his oral load caused tubular secretion that inflated the values.

Barrier, resistance and transporter details

Endothelial pores. The fenestrated endothelium has pores about 60 to 100 nm wide covering, in different estimates, a fifth to a half of its surface.

Resistance to water flow. In hydrodynamic models the endothelium adds little resistance to water flow; the basement membrane and the filtration slits, mainly the slit diaphragm, each supply roughly half.

Creatinine transporters. Proximal tubule cells take creatinine up through OAT2, OCT2 and OCT3 and release it through MATE1 and MATE2-K.

Who found tubuloglomerular feedback. A 2015 review credits Goormaghtigh with the juxtaglomerular idea (1937), Thurau and others with the feedback hypothesis in the early 1960s, and the Navar and Schnermann laboratories with demonstrating its role in autoregulation in the 1970s.